Short Answer

Both the model and the market overwhelmingly agree that the price of NVIDIA H200 compute will be above $2.35 by May 31, 2026, with only minor residual uncertainty.

1. Executive Verdict

  • HBM3e suppliers face severe constraints meeting NVIDIA H200 production demands.
  • Consistently high demand for H200 compute likely exerts upward price pressure.
  • Approved H200 sales to China may divert supply, elevating US cloud prices.
  • AMD MI300-series market share gains may influence NVIDIA H200 cloud rates.
  • Nebius announced plans to increase cloud GPU rental pricing effective June 1.

Who Wins and Why

Outcome Market Model Why
Outcome — — Insufficient data

Current Context

As of May 2026, NVIDIA H200 compute pricing varies across purchase and rental models. The purchase price for a single NVIDIA H200 GPU typically ranges from $30,000 to $45,000, influenced by form factor (PCIe versus SXM) and the chosen supplier [^][^]. Complete 8-GPU HGX H200 server systems are priced between $320,000 and $600,000, encompassing costs for integration, networking components, and memory-heavy infrastructure [^][^]. For cloud-based compute, on-demand rental rates for H200 GPUs generally fall between $2.50 and $10.60 per GPU-hour, with prices significantly differing by provider, commitment level, and regional availability [^][^][^][^].
Supply chain constraints and geopolitical factors influence H200 availability and cost. Ongoing bottlenecks in the supply chain, particularly regarding high-bandwidth memory (HBM3e), continue to exert upward pressure on hardware costs [^][^]. This occurs while enterprise demand for artificial intelligence infrastructure remains strong throughout Q2 2026 [^][^]. Furthermore, recent U.S. regulatory approvals have authorized the sale of H200 chips to approximately 10 major Chinese enterprises [^][^]. However, actual shipments are currently pending amid continued geopolitical tensions and supply-chain scrutiny [^][^].

2. Market Behavior & Price Dynamics

Historical Price (Probability)

Outcome probability
Date
This prediction market has exhibited a complete lack of price movement, maintaining a static 99.0% "YES" probability since its inception. The chart shows a sideways trend with no volatility, spikes, or drops. Because the price has remained unchanged from its starting point, it is not possible to identify any meaningful support or resistance levels. The market's stability is directly attributable to a total absence of trading activity.
The trading volume for this market has been zero contracts throughout its duration. This indicates a complete lack of participation or a very strong, unchallenged initial consensus. Without any trades, the market's price has not reacted to any external information, including the provided context on NVIDIA H200 pricing. The market's sentiment, as reflected by the 99.0% price, suggests an extremely high conviction in a "YES" outcome. However, this conviction is untested and represents the market's opening assessment rather than a price discovered through active trading and the incorporation of new information.

3. Market Data

View on Kalshi →

Contract Snapshot

The market resolves to Yes if the NVIDIA H200 compute value per hour is above $2.51 on May 31, 2026, as reported by Ornn; otherwise, it resolves to No. If no data is available by the May 31, 2026 expiration date, the market also resolves to No. Values are rounded to two decimal places, revisions after expiration are not accounted for, and the projected payout is June 1, 2026.

Available Contracts

Market options and current pricing

Outcome bucket Yes (price) No (price) Last trade probability

Market Discussion

As of May 2026, NVIDIA H200 compute rental prices are highly fragmented, commonly ranging from $2.50 to $7.00 per GPU-hour on-demand, with a broader reported range of $2.00 to $13.78 depending on the provider and commitment level [^][^][^][^][^]. Sustained high enterprise demand for memory-intensive AI workloads and recent U.S. government approval for H200 sales to major Chinese technology firms have maintained upward pressure on H200 availability and pricing in May 2026 [^][^][^]. This trend is consistent with the broader AI GPU rental market, which has seen a reported 20% year-to-date increase in H100 rental rates [^][^][^].

4. What is the likely impact on H200 cloud rental prices if the approved U.S. chip sales to Chinese enterprises are fulfilled before May 2026?

H200 rental price pressureUpward pressure expected due to US-China sales [^]
US H200 rental capacityExtremely tight as of early 2026 [^][^]
H200 hourly rate prediction$3.70/hr threshold confirmed for May 2026 [^][^][^][^][^]
Approved NVIDIA H200 chip sales to Chinese enterprises are likely to elevate US cloud rental prices. The fulfillment of these sales, especially before May 2026, is anticipated to create a "supply diversion" effect, which will reduce domestic US cloud capacity and consequently exert upward pressure on H200 hourly rental prices [^]. The US government had conditionally approved certain NVIDIA H200 exports to China in January 2026 [^].
The US market currently faces severely limited NVIDIA H200 rental capacity. As of early 2026, demand for AI inference workloads frequently surpasses available supply [^][^]. This ongoing situation contributes significantly to sustained or rising rental costs. Prediction markets for NVIDIA H200 compute prices by May 2026 indicate a strong upward trajectory, with resolutions confirming specific price thresholds, such as reaching $3.70 per hour, reflecting the severe supply constraints prevalent in the rental market [^][^][^][^][^].

5. What do Q4 2025 and Q1 2026 earnings reports from major HBM3e suppliers indicate about their ability to meet NVIDIA's H200 production demands?

HBM Demand Growth 2025130% year-on-year [^]
HBM Production Capacity 2026Fully booked [^][^][^]
DRAM Manufacturing Cycle18 to 36 months [^]
HBM3e suppliers face severe constraints meeting NVIDIA's H200 demands. Q4 2025 and Q1 2026 earnings reports from major HBM3e suppliers indicate their ability to meet NVIDIA's H200 production demands is severely limited by overwhelming demand and fully booked capacity through 2026 [^][^][^][^]. All three companies reported record revenues driven by AI memory products, yet their supply capacities are consistently falling short of customer requirements [^][^][^]. The overall outlook from these manufacturers highlights a significant imbalance between the surging demand for HBM3e, largely driven by AI accelerators like NVIDIA's H200, and the currently available supply [^].
Major HBM3e suppliers reported fully committed 2026 capacities. Micron Technology reported nearly $2 billion in HBM revenue in fiscal Q4 2025, primarily from HBM3e, and confirmed its entire HBM supply for 2026 is fully committed under binding contracts [^][^][^][^][^][^][^][^]. SK Hynix posted record-breaking earnings in Q4 2025 and Q1 2026, attributing this success to explosive AI demand for HBM, explicitly stating that customer demand for the next three years already surpasses their supply capacity [^][^][^][^][^][^]. Samsung Electronics demonstrated record consolidated revenue and operating profit in Q4 2025 and Q1 2026, propelled by a sharp increase in semiconductor sales, including HBM, and confirmed its HBM production capacity remains fully booked for the remainder of 2026 [^][^][^][^][^][^][^][^][^]. Samsung's memory chief specifically noted that current supply is falling well short of demand, with the gap expected to widen further into 2027 [^][^].
Long capital expenditure cycles exacerbate HBM3e supply shortages through 2026. The significant imbalance between HBM3e supply and demand is exacerbated by the long capital expenditure cycles inherent in DRAM manufacturing, which typically span 18 to 36 months, preventing rapid capacity expansion to meet sudden spikes in demand and ensuring supply remains constrained through 2026 [^]. HBM demand grew 130% year-on-year in 2025 and is projected to increase by 70% year-on-year in 2026 [^]. Additionally, reports suggest Samsung has encountered challenges with its 12-layer HBM3e stack qualification due to yield issues, which could potentially affect its ability to fully capitalize on market demand [^][^]. Furthermore, NVIDIA is reportedly evaluating additional H200 orders due to renewed demand from Chinese customers, a development that would further strain advanced packaging and wafer capacity, areas already identified as bottlenecks for high-bandwidth memory integration [^].

6. How does the total cost of ownership (TCO) for large-scale cloud providers compare between the NVIDIA H200 and its successor, the B200, as of Q1 2026?

B200 Memory Bandwidth8 TB/s [^][^][^][^][^]
H200 Memory Bandwidth4.8 TB/s [^][^][^][^][^]
H200 Hourly Rental Price (late May 2026)above $4.00 [^][^][^]
The NVIDIA B200 typically achieves a lower total cost of ownership (TCO) for large-scale training and inference, even with its higher per-GPU purchase price and operating cost compared to the H200 [^] [^] [^] [^] [^] . - Wecent">[^][^][^][^][^]. Conversely, the H200 proves more cost-effective for immediate, smaller-scale, or air-cooled deployments [^][^][^][^][^]. The B200's TCO advantage for massive models stems from its superior FP4 performance, greater memory bandwidth, and higher memory capacity, which lead to significant reductions in cost-per-token and training time [^][^][^][^][^]. Key specifications include the B200's memory bandwidth of 8 TB/s and capacity of 192GB, compared to the H200's 4.8 TB/s and 141GB, respectively [^][^][^][^][^]. The B200 also has a higher operating cost, indicated by a Thermal Design Power (TDP) of 1000W, while the H200's TDP is 700W [^][^][^][^][^].
Demand and supply constraints significantly impact H200 rental prices. As of late May 2026, prediction markets indicate that the hourly rental price for H200 compute has consistently exceeded $4.00, reflecting persistent high demand and supply constraints for Hopper-generation compute in the AI cloud market [^][^][^].

7. What public data sources or APIs track daily spot market prices for NVIDIA H200 instances across various cloud providers through Q2 2026?

Primary H200 IndexOrnnai H200 Index (dashboard.ornnai.com) [^][^][^][^]
Real-time Pricing APIsGPU Tracker (gputracker.dev/api-docs) and GetDeploying (getdeploying.com/help/api) [^][^]
Prediction Market ResolutionMany H200 contracts on Polymarket, Kalshi resolved using Ornnai Index, tracking rates until May 31, 2026 [^][^][^][^]
Public APIs offer real-time data for H200 spot market prices. Public data sources and APIs are available for tracking daily spot market prices for NVIDIA H200 instances across various cloud providers. GPU Tracker (gputracker.dev/api-docs) and GetDeploying (getdeploying.com/help/api) provide public APIs that aggregate real-time data for spot and on-demand GPU cloud pricing, encompassing H200 instances from multiple cloud providers [^][^].
The Ornnai H200 Index is crucial for H200 pricing data. The Ornnai H200 Index (dashboard.ornnai.com) serves as a primary data source for many prediction markets concerning NVIDIA H200 compute pricing. It provides finalized daily price points essential for contract resolution [^][^][^][^]. Prediction markets, including Polymarket and Kalshi, have utilized the Ornnai Index to resolve contracts through Q2 2026, which tracked whether H200 rental rates would meet specific thresholds by May 31, 2026 [^][^][^][^].

8. How might AMD's market share gains with its MI300-series accelerators in Q1 2026 affect cloud rental rates for the NVIDIA H200?

AMD Q1 2026 Data Center Revenue$5.8B (57% YoY increase) [^]
Median H200 On-Demand Rate~$3.89/GPU-hr [^]
H200 Prediction Market Outcome$3.70 on Apr 24, 2026 [^][^][^]
AMD's MI300 gains might influence NVIDIA H200 pricing. The MI300-series accelerators are expanding their market presence, evidenced by AMD's Q1 2026 Data Center revenue reaching $5.8 billion, a significant 57% year-over-year increase primarily driven by Instinct GPU shipments [^]. This growth introduces potential substitution effects that could reduce NVIDIA H200 pricing power in competitive cloud tiers [^][^][^].
NVIDIA H200 prices are stable despite competitive pressures. Despite the potential for substitution, current cloud GPU pricing trackers and prediction market outcomes indicate that robust H200 demand and ongoing supply tightness are expected to prevent prices from dropping dramatically into a lower band by late May 2026 [^][^][^][^]. H200 on-demand rental rates typically range around a few dollars per GPU-hour, with median tracked levels near ~$3.89/GPU-hr [^]. Supporting this stability, the prediction market contract for "Price of NVIDIA H200 compute by May 31, 2026?" on Coinbase Predictions resolved that at least some mid-tier price thresholds, such as the "$3.70" outcome, were met by April 24, 2026 [^][^][^].

9. What Could Change the Odds

Key Catalysts

A key bullish catalyst for H200 GPU rental prices is the reported approval by the US for H200 sales to about 10 major Chinese firms, which analysts viewed as supportive of sentiment and expected demand, as reported by Barchart mid-May 2026 [^].
Conversely, a potential bearish catalyst for demand emerged on May 21, 2026, when Nebius announced plans to increase on-demand pricing by an average ~29% and preemptible pricing by ~51% effective June 1 [^] . While this could drive up rental prices, it may also lead to customers throttling demand [^].
The market also watches specific price thresholds, such as those monitored by Polymarket's 'H200 hit by May 31' market, which resolves YES if the Ornnai H200 Index shows a finalized price at or beyond a specified threshold by May 31, 2026 [^] . Trading Odds & Predictions 2026 | Polymarket">[^]. For instance, one market for ↑ $4.00 resolved as YES on May 2, 2026 [^]. As of mid-May 2026, H200 cloud rental prices spanned approximately $1.19/hr to $10.60/hr, depending on the provider and region [^][^][^][^].

Key Dates & Catalysts

  • Strike Date: June 01, 2026
  • Expiration: June 08, 2026
  • Closes: June 01, 2026

10. Decision-Flipping Events

  • Trigger: A key bullish catalyst for H200 GPU rental prices is the reported approval by the US for H200 sales to about 10 major Chinese firms, which analysts viewed as supportive of sentiment and expected demand, as reported by Barchart mid-May 2026 [^] .
  • Trigger: Conversely, a potential bearish catalyst for demand emerged on May 21, 2026, when Nebius announced plans to increase on-demand pricing by an average ~29% and preemptible pricing by ~51% effective June 1 [^] .
  • Trigger: While this could drive up rental prices, it may also lead to customers throttling demand [^] .
  • Trigger: The market also watches specific price thresholds, such as those monitored by Polymarket's 'H200 hit by May 31' market, which resolves YES if the Ornnai H200 Index shows a finalized price at or beyond a specified threshold by May 31, 2026 [^] .

12. Historical Resolutions

Historical Resolutions: 20 markets in this series

Outcomes: 20 resolved YES, 0 resolved NO

Recent resolutions:

  • KXH200MON-26APR30-2.740: YES (May 01, 2026)
  • KXH200MON-26APR30-2.730: YES (May 01, 2026)
  • KXH200MON-26APR30-2.720: YES (May 01, 2026)
  • KXH200MON-26APR30-2.710: YES (May 01, 2026)
  • KXH200MON-26APR30-2.700: YES (May 01, 2026)